Optimized dispense recipes and 20nm filtration for reducing resist defects
06/01/2004
Point-of-use chemical filtration for 193nm photoresists and bottom antireflective coating (BARC) formulations has become a crucial issue as semiconductor manufacturers struggle with defect densities in next-generation processes. In this article, collaboration and tests demonstrate how 20nm filtration, combined with optimized resist-pump setup and dispense recipes, can significantly reduce defects in 193nm lithography.
Defect reduction by applying finer filtration to the various wafer processes has been a standard procedure in advancing semiconductor technologies, such as etch, CMP, and cleaning steps. The move to finer filtration in photolithography has caused some concern, however, because of its potential to affect CD control with polymer shear and surfactant stripping in photoresist materials. During the transition to ArF 193nm photoresist, initial installations focused on traditional metrics such as exposure latitude, depth of focus, CD, post-exposure bake temperature sensitivity, resist profile, and process windows. The need for reduced defect density was clearly identified during early implementation of these resists [1].
Original efforts in 1999 demonstrated that finer filtration of 193nm photoresist materials would not adversely affect these metrics [2]; however, these investigations did not address the issue of whether or not finer filtration in lithography could provide lower defect densities as it has in other semiconductor processing areas. In addition, defect densities of nonpatterned test wafers do not clearly represent defect densities observed in patterned wafers, after develop or after etch inspections. Microbridge defects associated with patterned 193nm resists [1] and defects associated with the corresponding BARC have been reported [3]. These problems have been identified as industry-wide phenomena, which warrants further investigation.
Filter construction
Typical point-of-use photoresist filters are constructed with a pleated membrane structure. The filtering membrane is sandwiched between appropriately designed support and drainage layers to provide a robust design. Hardware materials — including cage, core, and end caps — provide a rigid structure to facilitate handling.
LSI Logic Corp. and Pall Corp. have been working together to explore alternative materials and resist-pump process parameters for reduced defect densities and improved dispense recipes in 193nm lithography. The primary focus of this study was to examine the performance of sub-0.05µm rated filters to reduce various defects observed in BARC materials and 193nm resists. Multiple filter types were tested on a Tokyo Electron Ltd. Clean Track ACT8 tool utilizing the standard two-stage resist pumps. Lithographic performance of the filtered resist and BARC, and defect analysis of patterned wafers were performed. Optimized pump startup procedures and dispense recipes were also investigated to determine their effect on defect improvements. The track system used in this experiment was a standard production tool and was not modified from its original specifications.
The testing included filters with membranes constructed of various hydrophobic ultrahigh-molecular weight polyethylene (PE) and hydrophilic nylon 6,6 materials. These materials and construction techniques represent current industry standards; other construction techniques and materials may have the tendency to swell or soften in the highly polar solvents used in ArF photoresists. In addition, PE and nylon membranes offer superior pore-size control.
The joint tests involved installing a 0.02µm-rated nylon 6,6 filter and monitoring various defect densities, compared to existing PE media. Various defect types were identified along with the effects of the 20nm filtration, compared to standard industry filtration choices. The experiments required small capsule filters, incorporated in a standard dual-stage pump system. Adoption of 20nm nylon 6,6 filter capsules into the existing dual-stage pump was accomplished with the use of a manifold, allowing alternative capsules in the pump system.
193nm chemistries
The implementation of acrylate-based 193nm photoresists has brought with it widespread incidences of microbridge defects, as shown in Fig. 1. These defects have been reported in several recent papers [3] and have been observed in different formulations from various manufacturers. These resists, which utilize chemical amplification to greatly enhance photosensitivity, are based on a deblocking reaction catalyzed by photogenerated acid in exposed areas. This results in a solubility difference between the exposed and unexposed regions of the photoresist.
 Figure 1. Typical trench bridging defects, which have become a concern with the use of acrylate-based 193nm photoresists [3]. |
Contributing to this defectivity seems to be the various functional groups joined onto the acrylic polymer backbone, providing etch resistance, acid decomposition, adhesion promotion, and hydrophilicity. The naturally hydrophilic nylon 6,6 membrane filtration materials chosen for this installation have shown an affinity for reducing these defects [3]. It is hypothesized that the nylon 6,6 membrane appears as a polar surface in the 193nm solvent, which aids in the removal of low-solubility oligomers of nonfunctioning polymer components. The nylon 6,6 materials also provide a naturally hydrophilic pore structure, which eliminates phobic sites that can be commonplace with PE and surface-modified membranes. The nylon 6,6 membrane performance indicates an absorptive affinity for these undesired defects in addition to its superior sieving characteristics.
BARC materials are organic polymer films or inorganic films (IARC) with absorptivities and refractive indexes that match the photoresist material at the wavelength employed. Organic films are spun onto the wafer to suppress reflections from the substrate.
Performing the tests
Our testing involved developing a short loop of experimentation to determine the mechanism of a blocked poly defect. This defect is approximately 0.10–0.25µm in size and cannot be detected through standard inline defect monitoring at poly mask and poly etch. The defect is also highlighted only after spacer nitride etch. Because the defect appears after nitride etch, we theorize that it is initially a much smaller defect acting as a micro mask and becomes larger after the etch process. This defect type has been reported throughout the industry, and has evaded a proper solution due to its initial small size.
 Figure 2. Blocked poly defects. |
Testing found that replacing a standard BARC 0.1µm filter with a 0.02µm nylon 6,6 filter has greatly reduced this defect. The joint-testing effort also evaluated a polyethylene 0.02µm filter. Although the PE filter reduced the number of defects, the 0.02µm nylon 6,6 filter was 57% more efficient at reducing the blocked poly defect. Figure 2 shows the reduction benefits observed with the nylon 6,6 filter.
 Figure. 3 Typical "cone" defect. |
We believe one source of defects found after trench etch comes from impurities in the BARC material. These are commonly referred to as "cone" defects (Fig. 3). Significant defect reduction was observed when tests were switched to the 0.02µm nylon 6,6 filter. Total defect counts were reduced by 35%. New developable BARCs are being investigated to further reduce these defects. Figure 4 shows cone defects at island (isolated areas) with a track system designated "TRK1," utilizing 0.1µm PE filters on a deep-ultraviolet (DUV) BARC material, while TRK2 adopted the 0.02µm nylon 6,6 filter.
 Figure 4. Examples of cone defects at "island" (isolated areas). TRK1 uses 0.1??m PE filters on a DUV BARC material, while TRK2 adopted the 0.02??m nylon 6,6 filter. |
One defect type seen after poly etch is poly-blocked etch defects. An example of the poly-blocked etch defect is shown in Fig. 5. The use of a 0.02µm nylon 6,6 filter reduced defect counts by 23% over the standard 0.1µm PE filter.
 Figure 5. Typical poly-blocked etch defects: a) optical and b) SEM images. |
Test results showed that 0.03µm PE resist filters seemed to have little impact at island. However, 0.02µm nylon 6,6 BARC filters reduced defects by 35% for island and 23% for poly. New materials are being evaluated by photo to reduce cone defects even further. For metal/via structures, the nylon 6,6 resist filters have demonstrated a significant impact on defect reduction at metals by 17% at FI. Lowering pump filtration rates to 0.1ml/sec also shows a reduction of defects at via (50%) and metals (50%).
Based on test results, a strong correlation appears to exist between the implementation of the hydrophilic 0.02µm nylon 6,6 filters and reduction in defect densities. Although the 0.03µm PE filter has reduced the defects, the test showed that it is not good enough. Moving to a 0.02µm nylon 6,6 filter further reduced this defect at an efficiency 57% higher than the PE filter.
Pump issues
During the evaluation, we determined that finer filtration not only was possible, but a significant reduction in defect densities was also observed while utilizing a two-stage pump system designed to separate the filtration stage from the dispense stage. This design is aimed at allowing reduced filtration rates for minimized gel extrusion and filter pressure drop while maintaining dispense repeatability. The dispense stage and filtration stage do not appear to be completely independent or isolated, however, based upon test results.
An increase in pressure drop across smaller pore-size filters caused problems during the two-stage pump initialization subroutine. When dispensing in maintenance mode, very low defect levels were achieved. Once the system was switched to production mode and initialized, however, bubble formation was observed at the pump output line and at the dispense nozzle tip. Subsequent dispenses showed a high level of microbubble formation on inspected patterned wafers.
We hypothesized that during the transition from initialization mode to dispense mode a negative pressure was developed between the feed pump and dispense pump stages. This was verified by monitoring the internal pressure of the dispense stage during initialization and dispense modes with the lowest possible filtration rate selected. Raw data indicated a high negative-pressure drop during pump initialization (Fig. 6).
 Figure 6. Monitoring data indicates high negative-pressure drop during pump initialization. |
Negative pressure was also observed from the dispense stage to the dispense valve as shown by the formation of aspirated bubbles at the nozzle tip during the start of the first dispense, after initialization. This could be addressed by increasing the dispense-valve open delay time, but this would also compromise normal dispense condition setup. A pump-system software patch was implemented to reduce negative pressure to acceptable levels through modifications to valve and pump timing cycles during the initialization subroutine. Subsequent testing revealed no further problems.
Conclusion
Finer filtration combined with the hydrophilic nylon 6,6 media has demonstrated some of the lowest defect densities seen in testing. To enable further reduction in critical dimensions with minimized defects, finer filtration is needed. The implementation of 20nm filtration in the BARC process clearly shows reductions in after-etch defects. Problems that were previously attributed to the etch process are now being related to BARC chemistries. It appears that these problems can be reduced. Finer filtration of 193nm chemistries is possible, accompanied by greatly reduced defects. Current pump capabilities, combined with efforts to reduce filter size and holdup volume in capsule filters, may be counterintuitive to defect-reduction strategies. A balance needs to be established for using existing two-stage pump systems while trying to adopt finer filtration technologies.
References
- C.M. Jones, C. Kallingal, M. Zawadzki, N. Jeewakhan, N. Kaviani, et al., "Micro Photocell Monitoring Finds the Killers," Yield Management Solutions, Summer 2003.
- B. Gotlinsky, J. Beach, M. Mesawich, "Measuring the Effects of Sub 0.1µm Filtration on 248nm Photoresist Performance," SPIE Microlithography, pp. 3999–4138, March 2000.
- D. Hall, B. Gotlinsky, M. Mesawich, "The Effectiveness of sub 50nm Filtration on Reduced Defectivity in Advance Lithography Applications," ARCH Interface, Sept. 2003.
Phong Do received his bachelors in physics from The College of the Holy Cross, Worcester, MA. He is a senior photo process development engineer at LSI Logic Corp., 23400 N.E. Glisan St., Gresham, OR 97030-8411; ph 503/618-4134, e-mail [email protected].
Joe Pender attended DeVry Institute of Technology and is a 26-year industry veteran. He is a photo equipment engineer on TEL track systems at LSI Logic.
Thomas R. Lehmann has served as a metrology equipment lab instructor for the US Air Force at the Advanced PMEL Laboratory, and as an instructor for advanced test measurement and diagnostic equipment at the Army Advanced Calibration Laboratory. Lehmann is involved in defect analysis and research into chemical filtration for advanced lithography processes at LSI Logic.
Barry Gotlinsky received his PhD in chemistry from the City U. of New York. He is VP of microelectronics support in the Scientific Laboratory Services Department of Pall Corp., 2200 Northern Blvd., East Hills, NY 11548; ph 516/801-9260, e-mail [email protected].
Michael Mesawich received his BS in physics from the State U. of New York at Stony Brook. He is VP of lithography at Pall Microelectronics.